Quantitative comparison of two-dimensional and three-dimensional strain measurement using MRI feature tracking in repair Fontan patients and normal child volunteers

Functional single ventricle (FSV) is a type of severe congenital heart disease (CHD) [1]. The first palliation surgery for FSV was described by Fontan and Baudet [2] and performed in 1971 to treat tricuspid atresia. In patients who have undergone this Fontan repair procedure, a single ventricle provides blood flow in series to the pulmonary and systemic circulation. Young survivors with Fontan circulation commonly develop ventricular dysfunction, which has been identified as a risk factor for mortality [3, 4]. Moreover, long-term follow-up studies over the last 20 years have used echocardiography (echo) to assess ventricular function after successful Fontan surgery in childhood and shown significant differences in regional deformation and ejection fraction [5‐7].

Cardiac magnetic resonance (CMR) has recently become the gold standard for measurement of ventricular function and myocardial motion [8]. CMR feature tracking (FT) has been adapted and applied to the standard CMR sequence (balanced steady-state free precession, b-SSFP) without additional sequences, unlike in myocardial tagging or displacement encoding with stimulated echoes (DENSE) imaging [9]. Strain assessment by both echo and CMR shows high intra-modality and modest inter-modality reproducibility [10]. At present, our study found that two-dimensional (2D) strain analysis is considered a suitable tool to detect early abnormalities of the ventricular myocardium [11]. However, ventricular function assessment is challenging in post-Fontan FSV patients due to the complex ventricular geometry. Some studies have indicated that three-dimensional (3D) strain analysis may overcome such geometry-dependent limitations of 2D strain analysis by referencing the intrinsic directions of deformation [12, 13]. To our knowledge, there are no previous studies assessing three-dimensional strain analysis of SSFP cine images in Fontan patients and normal volunteers [14]. Thus, the aim of this study is to explore the feasibility of 3D strain analysis with 2D cine CMR images that may be clinically useful in the assessment of post Fontan patients.

The Fontan procedure is the final step in the staged palliative repair of FSV [16], and children who have undergone this procedure are at risk for clinical complications. Fontan repair patients show comparatively good preservation of ventricular function in the intermediate term; however, ventricular function deteriorates in most patients over time [17]. Ventricular dysfunction attributed to long-term volume and pressure overload could have been more prominent in the post-Fontan operation by average five-years follow-up.

CMR-FT is a non-invasive modality for evaluating ventricular function in Fontan repair patients. It has been used to analyze ventricular function in CHD, particularly in cases of coarctation of the aorta [18] and repaired transposition of the great arteries [19]. In the previous publications, Satriano A et al. mentioned that 3D principal strain analysis from routine 2D cine CMR imaging is clinical feasible, highly reproducible, and shows strong correlations with conventional measures of strain [20]. 3D strain clinical research had been reported in recent years [21]. Considering the difficulties in ventricular function assessment in Fontan FSV repair patients due to the complex ventricular geometry, 3D strain analysis by FT may be a more accurate tool than 2D strain analysis [12]. In a study of 100 healthy subjects. Liu et al. [22] showed that the 3D global strain values ranged from 22 to 73 for radial strain, − 9 to − 20 for circumferential strain, and − 13 to − 23 for longitudinal strain. In our study, the 32 control group volunteers showed strain values within the reference range. In another study by André et al. [23] that used parabolic regression equations, no sex-related differences were found in the assessed strain values in children and adolescents. Our study showed significant age differences between the patient group and control group. There were no statistically significant CMR volumetric differences between the groups despite this age discrepancy, which might suggest that FSV patients are already demonstrating ventricular dilatation. All three of the 3D GLS, GCS and GRS values in the Fontan repair patients was significantly lower than those in the child volunteers. In contrast, only the 2D global longitudinal strain showed significant differences, which has been consistently reported in previous studies. 2D GLS has been the most valuable clinical marker for the assessment of myocardial disease [24]. We found that the 3D global strain was lower than the 2D strain in the patient group. We propose the following explanation for this finding: In 2D CMR-FT, the out-of-plane motion of one segment exaggerates the perceived degree of muscle shortening, thereby resulting in overestimation of myocardial movement. 3D CMR-FT based on the 3D incompressible model-based algorithm can overcome the assumption of cylindrical chamber architecture and therefore produces lower values [20]. 3D strain assessments may be useful for monitoring abnormal regional myocardial motion in patients with Fontan circulation. In our study, 3D RCS decrease was marked in the basal and apical segments (P < 0.001). it could be related to abnormal cardiac looping, which leads to hearts that lack helical fiber patterns. However, this theory needs further study.

The 2D strain algorithm fits a 2D incompressible deformable cylindrical model of the myocardium to individual image slices over the entire cardiac cycle. This model is assumed to be completely determined by control points placed on the middle curve of the endo and epi-boundaries of the myocardial wall. The model is based on the feature-tracked boundaries and the incompressibility constraint of the model. In contrast, the 3D strain algorithm uses a 3D deformable model generated by interpolating the tracked boundaries from the 2D algorithm. The surface interpolation is performed using both long- and short-axis image information [25]. The removal of constraints through pre-determined geometry-dependent directions of deformation may therefore, provide a more accurate and reproducible measure of Fontan repair cases. On the basis of these principles, the 3D regional strain offers advantages in analyzing non-cylindrical structure models. In our study, patients with Fontan repair included 21 cases with left ventricular morphology and 11 cases with right ventricular morphology. In order to quantify strain must standardize anatomically the myocardial regions. However, it is challenging in patients with single RV. Due to the complex ventricular geometry, this result is reinforced by the regional deformation analysis of the different levels [20]. This difference between 2D and 3D regional strain needs to be investigated further in a patient group.

Berganza et al. reported that the 3D GCS was associated with the indexed right ventricular EDV in patients with repaired tetralogy of Fallot [12]. We found a strong correlation between 3D GCS and indexed EDV, indexed ESV, EF in the Fontan repair patients. Meanwhile, the 3D strain showed better reproducibility than the 2D strain in our study. 3D GCS showed the best inter-observer agreement (ICC = 0.87, CoV = 5.15), followed by 3D GLS (ICC = 0.84, CoV = 5.36) as compared to 2D GCS (ICC 0.82, CoV = 5.73). Previous studies reported that CMR-FT–derived 2D GLS had the best intermodality agreement among the indices measured, followed by 2D GCS, whereas 2D GRS showed more degrees of divergence between CMR-FT and speckle tracking echocardiography [26, 27]. Our results consistently indicated the high clinical applicability of 3D GCS and GLS as most robust parameters. In our study, the coefficient of variation of 3D GRS was not good (Fig. 4). GRS represented strain throughout the entire myocardial wall from subepicardium to subendocardium and was consequently much more affected by through plane motion and complex diastolic and systolic twisting motion [28].

Our study was limited by its single-center design. The retrospective nature of the study also limited its population to those clinically referred for CMR, allowing for a potential selection bias. Due to the limitations of acquisition time, we collected SSFP cine images in the four-chamber, two-chamber, and short-axis views, and the three-chamber view was not included. Result had been affected on the 2D and 3D strain analysis without three-chamber view, this is also the limitation of retrospective study. There was evidence from previous studies that the problem of twisting motion and out of plane motion could affected using 3D strain compared with 2D algorithms. In addition, 3D images presented a substantially lower spatial and temporal resolution than their 2D counterpart. We inferred that these findings might be the reason why 3D GRS results are less stable and effective [29]. Furthermore, strain rate analysis was not included in our research because the temporal resolution of the CMR cine images was somewhat lower than what is commonly recommended for strain rate evaluation in children [24]. Previous literature reported that 3D GLS in healthy subjects had a remarkably lower values compared with 2D GLS, reference values of GLS using 2D speckle tracking analysis and 3D speckle tracking analysis was 21 and 19%, respectively. The current 3D FT software has a significant drawback to measure 3D GLS [30]. Further studies are needed to assess Fontan repair patients using 3D strain and 3D speckle-tracking echocardiography.

3D GCS is feasible, highly reproducible, and strongly correlated with conventional cardiac function measures. 3D GCS assessments may be useful for monitoring abnormal myocardial motion in patients with Fontan circulation.